I’ve looked into crop for ITS clustering (where a hard % cutoff is even more theoretically questionable than in 16S)
https://code.google.com/p/crop-tingchenlab/